China's larch stock volume estimation using Sentinel-2 and LiDAR data

Forest Stock Volume (FSV) is one of the key indicators in forestry resource investigation and management on local, regional, and national scales. Limited by the saturation problems of optical satellite remote-sensing imagery in the retrieving of stock volume, and the high cost of Light Detection And Ranging (LiDAR) data, it is still challenging to estimate FSV in a large area using single-sensor remote-sensing data. In this paper, a method integrated multispectral satellite imagery and LiDAR data was developed to map stock volume in a large area. A random forest model was adopted to estimate the stock volume of larch forest in China based on the training samples from the Airborne Laser Scanning (ALS)-derived stock volume and corresponding Sentinel-2 imagery. Validation using National Forest Inventory (NFI) data, ALS-derived stock volume and ground investigation data demonstrated that the estimated stock volume had a high accuracy (R-2 = 0.59, RMSE = 59.69 m(3)/ha, MD = 39.96 m(3)/ha when validated with NFI data; R-2 ranged from 0.77 to 0.85, RMSE ranged from 38.68 m(3)/ha to 67.38 m(3)/ha, MD ranged from 24.90 m(3)/ha to 37.27 m(3)/ha when validated with ALS stock volume; R-2 = 0.42, RMSE = 79.10 m(3)/ha, MD = 62.06 m(3)/ha when validated with field investigation data). Results of this paper indicated the applicability of estimating stock volume of larch forest in a large area by combining Sentinel-2 data and airborne LiDAR data.